U.S. patent application number 10/583664 was filed with the patent office on 2009-07-09 for apparatus, and associated method, for detecting packets.
This patent application is currently assigned to NOKIA CORPORATION. Invention is credited to Paolo Priotti.
Application Number | 20090175362 10/583664 |
Document ID | / |
Family ID | 34793605 |
Filed Date | 2009-07-09 |
United States Patent
Application |
20090175362 |
Kind Code |
A1 |
Priotti; Paolo |
July 9, 2009 |
APPARATUS, AND ASSOCIATED METHOD, FOR DETECTING PACKETS
Abstract
In a robust system and method for detecting packets in SISO and
MIMO broadband multicarrier transmission, a packet detector
computes continuously the sum of the moduli of the power-normalized
auto-correlations of a sequence of received signals (204), tuned on
the periodicity of the training symbols. When the power-normalized
auto-correlation exceeds a first predetermined threshold (206),
then the maximum value of the sum of the moduli of all the
cross-correlations between the received signals and the M aperiodic
sequences is computed in a given time window (208), sliding in time
from -c.sub.0 to c.sub.0. A packet is identified as received (212)
when the maximum value of the sum of the moduli of the
cross-correlations exceeds a second predetermined threshold
(210).
Inventors: |
Priotti; Paolo; (Turin,
IT) |
Correspondence
Address: |
BANNER & WITCOFF, LTD.
1100 13th STREET, N.W., SUITE 1200
WASHINGTON
DC
20005-4051
US
|
Assignee: |
NOKIA CORPORATION
Espoo
FI
|
Family ID: |
34793605 |
Appl. No.: |
10/583664 |
Filed: |
December 29, 2003 |
PCT Filed: |
December 29, 2003 |
PCT NO: |
PCT/US03/41506 |
371 Date: |
October 3, 2008 |
Current U.S.
Class: |
375/260 |
Current CPC
Class: |
H04B 7/04 20130101; H04L
27/2656 20130101; H04W 28/18 20130101; H04L 27/2675 20130101 |
Class at
Publication: |
375/260 |
International
Class: |
H04L 27/28 20060101
H04L027/28 |
Claims
1. A method for detecting a packet in multiple signals received in
parallel from at least two transmission antennas, the method being
characterized by steps of: computing the power-normalized
auto-correlations of the received signals tuned on the periodicity
of short training symbols, said short training symbols having time
periods shorter than one OFDM symbol; determining whether the sum
of the moduli of the power-normalized, auto-correlation exceeds a
first predetermined threshold; upon a determination that the sum of
the moduli of a power-normalized, auto-correlation exceeds the
first predetermined threshold, computing the cross-correlations
between the received signals and the aperiodic sequences in a
selected time window, said time window sliding in time for a
quantity larger than the variance of the first predetermined
threshold crossing instant; computing the maximum value of the sum
of the moduli of the cross-correlations; determining whether the
maximum value of the sum of the moduli of the cross-correlations
exceeds a second predetermined threshold; and upon a determination
that the maximum value of the sum of the moduli of the
cross-correlations exceeds a second predetermined threshold,
identifying a packet as received.
2. The method of claim 1 further characterized in that, upon a
determination that the sum of the moduli of a power-normalized,
auto-correlation does not exceed the first predetermined threshold,
repeating the step of computing the power-normalized
auto-correlations of next received signals tuned on the periodicity
of short training symbols.
3. The method of claim 1 further characterized in that the short
training symbols constitute a first part of a training sequence
comprising at least one OFDM symbol having a time-periodic
component effective for packet detection and coarse frequency
offset correction.
4. The method of claim 1 further characterized in that the short
training symbols constitute the first part of a training sequence,
and wherein the part of the training sequence used for packet
detection is contained in the first at least one OFDM symbol of the
whole training sequence.
5. The method of claim 1 further characterized in that the periodic
portion of the received signals is substantially orthogonal between
said at least two transmission antennas.
6. The method of claim 1 further characterized in that the first
part of the training sequence is further characterized by a
non-periodic component effective for rejecting interferers.
7. The method of claim 1 further characterized in that the first
part of the training sequence is further characterized by a
non-periodic component effective for rejecting interferers, said
interferer comprising a DC component.
8. The method of claim 1 wherein the first part of the training
sequence is further characterized by a non-periodic component in
the first OFDM symbol, and is effective for rejecting
interferers.
9. The method of claim 1 further characterized in that the first
part of the training sequence is further characterized by a
non-periodic component in the first OFDM symbol, and is effective
for rejecting interferers, said interferers including a DC
component.
10. The method of claim 1 further characterized in that the
non-periodic component is orthogonal between said at least two
transmission antennas.
11. An apparatus for detecting a packet in multiple signals
received in parallel from at least two transmission antennas, the
apparatus being characterized by: a first computational portion
configured for computing the power-normalized auto-correlations of
the received signals tuned on the periodicity of short training
symbols, said short training symbols having time periods shorter
than one OFDM symbol; a first comparator configured for determining
when the sum of the moduli of the power-normalized,
auto-correlation exceeds a first predetermined threshold; a second
computational portion configured, upon a determination that the sum
of the moduli of a power-normalized, auto-correlation exceeds the
first predetermined threshold, for computing the cross-correlations
between the received signals and the aperiodic sequences in a
selected time window, said time window sliding in time for a
quantity larger than the variance of the first predetermined
threshold crossing instant; and a second comparator configured for
computing the maximum value of the sum of the moduli of the
cross-correlations, and for determining when the maximum value of
the sum of the moduli of the cross-correlations exceeds a second
predetermined threshold, and upon a determination that the maximum
value of the sum of the moduli of the cross-correlations exceeds a
second predetermined threshold, for identifying a packet as
received.
12. The apparatus of claim 11, wherein said first computational
portion is further characterized, upon a determination that the sum
of the moduli of a power-normalized, auto-correlation does not
exceed the first predetermined threshold, as computing the
power-normalized auto-correlations of next received signals tuned
on the periodicity of short training symbols.
13. The apparatus of claim 11 further characterized in that the
short training symbols constitute a first part of a training
sequence comprising at least one OFDM symbol having a time-periodic
component effective for packet detection and coarse frequency
offset correction.
14. The apparatus of claim 11 further characterized in that the
short training symbols constitute the first part of a training
sequence, and wherein the part of the training sequence used for
packet detection is contained in the first at least one OFDM symbol
of the whole training sequence.
15. The apparatus of claim 11 further characterized in that the
periodic portion of the received signals is substantially
orthogonal between said at least two transmission antennas.
16. The apparatus of claim 11 further characterized in that the
first part of the training sequence is further characterized by a
non-periodic component effective for rejecting interferers.
17. The apparatus of claim 11 further characterized in that the
first part of the training sequence is further characterized by a
non-periodic component effective for rejecting interferers, said
interferer comprising a DC component.
18. The apparatus of claim 11 wherein the first part of the
training sequence is further characterized by a non-periodic
component in the first OFDM symbol, and is effective for rejecting
interferers.
19. The apparatus of claim 11 further characterized in that the
first part of the training sequence is further characterized by a
non-periodic component in the first OFDM symbol, and is effective
for rejecting interferers, said interferers including a DC
component.
20. The apparatus of claim 11 further characterized in that the
non-periodic component is orthogonal between said at least two
transmission antennas.
Description
TECHNICAL FIELD
[0001] The present invention relates in general to the field of
wireless transmission and, more specifically, to broadband
multicarrier transmission links, and still more specifically, to a
robust system and method and training sequence for detecting
packets in SISO and MIMO broadband multicarrier transmission.
BACKGROUND
[0002] In packet-based systems, where the arrival timing of a
packet is not known a-priori, there is a need to detect an incoming
packet to trigger events in the receiver, such as the
synchronization chain. Among the W-LAN (wireless local area
network) systems developed in recent years, there are systems that
have a wide bandwidth ("BW") and are based on multicarrier
modulation. These systems have a system timing that is very fast in
absolute terms, and the same is anticipated with respect to
next-generation cellular systems, where sampling rates will be on
the order of many tens of Msps. In light of this background, it is
necessary to have algorithms for packet detection that are (1) of
limited computational complexity to allow fast processing, (2) not
prone to false alarms, (3) that detect promptly packets even at the
lower edge of the operating SNR (signal-to-noise ratio) region, and
(4) if based on a training sequence, then the same sequence has to
be BW-efficient and have low PAPR (peak-to-average power
ratio).
[0003] To assure real-time processing, the packet detection
algorithm must be relatively simple, and for this reason it is
usually based on the auto-correlation of a given segment of the
incoming signal. This guarantees better performance than algorithms
that simply monitor the incoming signal energy. In particular,
recent implementations seem to favor the use of a short repetitive
pattern, used both for packet detection and for coarse frequency
offset compensation. This short repetitive pattern is often
referred to as utilizing "short training symbols" to indicate that
one time period of the sequence is shorter than one OFDM
(orthogonal frequency division multiplexing) symbol.
[0004] There are, however, drawbacks associated with using a short
repetitive pattern, in that the algorithm tends to recognize as an
incoming packet every repetitive noise pattern. In particular, the
algorithms used with a short repetitive pattern are prone to false
alarms when (1) there is a DC component in the input (interpreted
as repetitive pattern) as has been the case with IEEE802.11a HW
implementations, and (2) there is co-channel interference.
[0005] There is, therefore, a need for algorithms that can
discriminate more effectively between packets and noise or
interference. It should be noted that simple digital filtering can
be applied to block a DC (direct current) component, but this can
degrade somewhat the incoming signal. It should also be noted that,
for next-generation cellular systems, co-channel interference
problems in the case of frequency reuse factor 1 are expected to be
more relevant than for W-LANs, where in many cases you do not have
an adjacent cell directly interfering in your operation area.
SUMMARY
[0006] The present invention accordingly provides for a packet
detection system that makes use of combined auto-correlation of the
received signal and cross-correlation of the signal with a portion
of the training sequence. The algorithm is based on a training
sequence where the first part (e.g., preferably one or two OFDM
symbols) contains a time-periodic component that can be used for
both packet detection and coarse frequency offset correction. This
periodic part is orthogonal between the various TX (transmission)
antennas. The first part of the training sequence contains also a
non-periodic component (typically in the very first OFDM symbol)
used for rejecting interferers such as a DC component.
Additionally, the non-periodic component is chosen to be orthogonal
between different antennas.
[0007] The packet detector computes continuously the
auto-correlation of the combined RX (received) signal, tuned on the
periodicity of the short training symbols. When the
power-normalized auto-correlation exceeds a first predetermined
threshold, then the cross-correlations between the received signal
and all the non-periodic components are computed in a given time
window, short (e.g., 2K is small in comparison to LN) and close
(e.g., the aperiodic sequence has been inserted in the first OFDM
symbol) to the head of the packet. This process is preferably
repeated, sliding time-wise for a quantity larger than the variance
of the instant when the first threshold is crossed. A packet is
considered received when the maximum value of the sum of the moduli
of the cross-correlations exceeds a second predetermined
threshold.
[0008] The complexity of the algorithm is kept low by limiting the
length of correlation windows. For example, the length of a
correlation window may be less than a standard symbol timing
recovery algorithm for MIMO. Thus, if the symbol timing can be
found in real time, the packet detection can also be performed in
real time.
[0009] As a result of using two independent thresholds, the
algorithm is more robust than algorithms based on one decision
variable only (such as auto-correlation). As described in further
detail below, the use of two decision variables permits, in most
cases, discrimination between a desired packet and noise.
BRIEF DESCRIPTION OF THE DRAWINGS
[0010] For a more complete understanding of the present invention,
and the advantages thereof, reference is now made to the following
descriptions taken in conjunction with the accompanying drawings,
in which:
[0011] FIG. 1 presents a block diagram of the proposed
algorithm;
[0012] FIG. 2 presents a flow chart showing steps of operation of
the invention;
[0013] FIG. 3 presents a simulation environment used for testing
the proposed packet detection algorithm;
[0014] FIG. 4 presents the structure of the training sequence
transmitted from a single antenna;
[0015] FIGS. 5A, 5B, and 5C present the effect on the
auto-correlation for different placements of the aperiodic
portion;
[0016] FIGS. 6A, 6B, 6C, 6D, and 6E present the changes in the
autocorrelation in the presence of coloured co-channel interferers,
and the cross-correlation in the same conditions;
[0017] FIG. 7 presents the complete structure of the training
sequence used in the implementation; and
[0018] FIG. 8 presents a comparison of the performance of the
proposed system with the auto-correlation based approach.
DETAILED DESCRIPTION
[0019] In the following discussion, numerous specific details are
set forth to provide a thorough understanding of the present
invention. In some instances, well-known elements have been
illustrated in schematic or block diagram form in order not to
obscure the present invention in unnecessary detail. Additionally,
for the most part, details concerning wireless transmission,
broadband multicarrier transmission links, packet-based systems,
and the like have been omitted, except insofar as necessary to
describe the present invention, inasmuch as such details are not
considered necessary to obtain a complete understanding of the
present invention, and are considered to be within the skills of
persons of ordinary skill in the relevant art.
[0020] It is noted that, unless indicated otherwise, all functions
described herein are performed by a data processor in accordance
with code. As used herein, the term "data processor" shall include
and be used to refer to any one or more of a microprocessor, a
microcontroller, an application-specific integrated circuit (ASIC),
a device (e.g., a personal digital assistant (PDA), a mobile
telephone, or the like), an electronic data processor (EDP), a
computer, a personal computer (PC) and/or the like. Furthermore, as
used herein, the term "code" shall include and be used to refer to
any one or more of program code, software, integrated circuits,
read-only memory (ROM), and/or the like, effective for instructing
the data processor how to perform such functions. Still further, it
is considered that the design, development, and implementation
details of all such code would be apparent to a person having
ordinary skill in the art based upon a review of the present
description of the invention.
[0021] The invention can operate on both SC (single-carrier) and MC
(Multi-Carrier) systems, and in both SISO (single-input,
single-output) and MIMO (multiple-input, multiple-output) cases
(plus combinations such as MISO and the like). Simulations have
been conducted using an MC-MIMO system and, accordingly, the
notation used herein refers to a system having N subcarriers, M TX
antennas, and P RX antennas. In particular, the present system can
bring benefit in future cellular communication systems, that are
likely to be MC-MIMO. However, the following paragraphs apply fully
to SC SISO and MIMO systems as well.
[0022] Let the OFDM signal at the m-th TX antenna be:
x m ( t ) = 1 N n = 0 N - 1 X m ( n ) j 2 .pi. nt / N , m = 1 M ( 1
) ##EQU00001##
[0023] Then let the received signal at the p-th RX antenna be:
r p ( t ) = m = 1 M l = 1 .DELTA. - 1 c l mp ( t ) x m ( t - l ) +
v p ( t ) , p = 1 P ( 2 ) ##EQU00002##
where .DELTA. represents the maximum delay spread of the channel
(the time unit is the sampling time), c.sub.l.sup.mp is the
time-variant coefficient for the l-th tap in MIMO sub-channel mp,
and the first tap is placed in the time origin. .nu..sub.p is an
additive noise contribution.
[0024] The problem may be formulated as one of detecting the
presence of the desired signal r.sub.p with a high reliability,
without detecting other signals (e.g., thermal noise, colored
interferers, and the like). The detection of the desired signal
preferably occurs within a given time range from the start instant
of said signal.
[0025] The performance of a packet detection algorithm can be
decided by estimating the probability P.sub.FA of a false alarm
(also referred to as false detection), and the probability of
no-detection P.sub.ND (also referred to as miss-detection).
P.sub.FA is calculated as the probability that an interfering
signal or noise will be interpreted as a desired packet, and will
trigger the receiver on, when no data can be received, causing a
waste of power and, in a worst case, the loss of a desired packet.
P.sub.ND is calculated as the probability that a desired packet is
not recognized by the detection algorithm and, as such, is
ignored.
[0026] Most implementations of packet detection use either the
energy of the received signal or its auto-correlation as decision
parameter. When one of these quantities exceeds a given threshold,
a packet is considered received. However, these implementations
based on a single decision parameter suffer from the following
problem. If the threshold is too low, P.sub.FA becomes too high,
but if the threshold is too high, P.sub.ND becomes too high. The
threshold value is chosen as a compromise between these two
opposite values. The threshold value can, as such, be optimized to
ensure that jointly P.sub.FA and P.sub.ND are low on relative terms
for a given environment, but for changing SNR and interference
conditions, the performance can degrade, as discussed further
below. It should be noted that, in MIMO systems, the use of space
diversity guarantees a more constant received power of the combined
signal, in comparison with a SISO receiver, and, as such,
auto-correlation properties are improved, but the algorithm remains
equally weak to colored interferers.
[0027] A more robust solution is proposed herein, wherein one more
degree of freedom is provided to the detection algorithm by using
two different decision parameters. In this way, two thresholds can
be chosen separately and bring a higher degree of optimization. The
present invention uses auto-correlation and cross-correlation as
decision variables because the implementation has low
complexity.
[0028] The training sequence proposed here is given by the
combination of a time-periodic signal and an aperiodic signal. It
is compatible with the implementation for carrier frequency offset
synchronization disclosed in co-pending U.S. patent application
Ser. No. 10/646,524, entitled "FREQUENCY SYNCHRONIZATION OF MIMO
OFDM SYSTEMS WITH FREQUENCY-SELECTIVE WEIGHTING" and filed on Aug.
22, 2003. The overlapped aperiodic signal does not hinder correct
operation of frequency offset synchronization.
[0029] The first two OFDM symbols contained in the training
sequence are defined in the time domain as:
x ~ m ( t ) = k = 0 S .xi. - 1 C m D ( t - kD ) + C m GD ( t -
.tau. ) , t = 0 S N - 1 , .tau. < N - G D ( 3 ) ##EQU00003##
where D=N/.xi. represents the time period of the periodic
component, C.sub.m.sup.D is a pseudo-random sequence of length D
used to build the periodic component, {hacek over (C)}.sub.m.sup.GD
is a random sequence of length G D that represents the aperiodic
component. C.sub.m.sup.D and {hacek over (C)}.sub.m.sup.GD are
chosen so that a pseudo-orthogonality condition between different
TX antennas is achieved:
.A-inverted. m ' , m '' .di-elect cons. { 1 M } , t = 0 SN - 1 x ~
m ' ( t ) x ~ m '' * ( t + k ) = { .alpha. m if m ' = m '' = m AND
K = 0 .apprxeq. 0 otherwise . ( 4 ) ##EQU00004##
[0030] The pseudo-orthogonality condition is broader than the
orthogonality condition and, as such, is inclusive of the cases in
which the random sequences used to build the training sequence
comprise orthogonal codes, such as Walsh-Hadamard codes.
[0031] Giving a definition of {tilde over (x)}.sub.m in the time
domain, PAPR may be readily controlled. Alternatively, a definition
in the frequency domain is also possible, wherein the periodic part
uses only one every .xi. subcarriers. This possibility is not
explored further here, but rather, a definition in the time domain
alternative to (3) is given, for the case where a PAPR of 0 dB or
close to 0 dB is required (this is the case when the training
sequence has to be transmitted with a boosted power level, but
signal distortion is unacceptable).
x ~ m ( t ) = w ( t ) k = 0 S .xi. - 1 C m D ( t - kD ) + C m GD (
t - .tau. ) , t = 0 S N - 1 , .tau. < N - G D Where w ( t ) = {
1 / 2 if .tau. .ltoreq. t < .tau. + G D 1 otherwise ( 3 ` )
##EQU00005##
and in this case {hacek over (C)}.sub.m.sup.GD is chosen to have an
average power level 3 dB lower than C.sub.m.sup.D.
[0032] The part of the training sequence used by packet detection
is preferably contained in the first one or two OFDM symbols of the
whole training sequence to ensure that packet detection itself can
operate as soon as the head of the packet has been received. For
this reason the value of .tau. will preferably be chosen around N/2
and a choice of 1.ltoreq.G.ltoreq.2 is also appropriate.
[0033] With reference to FIG. 1 of the drawings, the reference
numeral 100 generally designates a packet detection apparatus of a
receiver embodying features of the present invention. The packet
detection apparatus 100 includes a first computational portion 106
coupled for receiving P signals from one or more receiver (RX)
antennas, exemplified herein by two antennas 102. The first
computational portion 106 is configured for computing continuously
the power-normalized, auto-correlation of the P signals received
from the RX antennas 102. More specifically, the functionality of
the first computational portion 106 is achieved utilizing, for each
RX antenna 102, a portion 106a configured for computing the
absolute value of auto-correlation of samples 0.about.LN received
from a respective RX antenna 102, and a portion 106b configured for
computing the sliding window for average power of samples
0.about.LN received from a respective RX antenna 102. Depending on
receiver system design, a backoff and enable portion 106c may
optionally be provided for disabling the entire packet detection
apparatus 100 for a given time interval, after a packet has been
recognized as detected.
[0034] A first comparator 108 is connected for receiving the
computed power-normalized, auto-correlation of the received signals
and for generating a trigger signal to a second computational
portion 110, discussed below, when the power-normalized,
auto-correlation exceeds a first predetermined threshold
.phi..sub.1, such as exemplified in the implementation discussed
below with reference to FIGS. 3-8, wherein .phi..sub.1=0.4M, M
representing the number of RX antennas 102.
[0035] The second computational portion 110 is coupled, upon
receipt of a trigger signal from the first comparator 108, for
receiving P signals also from the one or more RX antennas,
exemplified by the two antennas 102, and for cross-correlating the
received P signals with the aperiodic portion of the training
sequence. More specifically, the functionality of the second
computational portion 110 is achieved utilizing, for each RX
antenna, a portion 110a configured for computing the absolute value
of cross-correlation samples of samples .tau.-K.about..tau.+K
(performed with up to M different aperiodic sequences and sliding
in time from -c.sub.0 to c.sub.0) received from a respective
antenna, and a portion 110b configured for computing the sliding
window for average power of samples .tau.-K.about..tau.+K received
from a respective antenna.
[0036] A second comparator 112 is coupled to the second
computational portion 110 for determining whether the maximum value
of the sum of the moduli of the cross-correlations exceed a second
predetermined threshold .phi..sub.2, and if it does, then for
generating a signal indicating that a packet has been received and
detected. The generated signal may transmitted to other components
of a receiver suitable to trigger further events in the receiver,
such as symbol timing recovery, carrier frequency synchronization,
an FFT, and/or the like, depending on how the invention is
implemented.
[0037] Operation of the algorithm of the packet detection apparatus
100 in accordance with the present invention is exemplified by a
flow chart 200 shown in FIG. 2. Accordingly, in step 202, P signals
are received in parallel from the RX antennas 102. In step 204, the
first portion of the algorithm is executed by the first
computational portion 106, to continuously calculate the
power-normalized auto-correlation .PSI..sub.p(k), that is always
active, except during the reception of the payload and a possible
backoff time after that.
.PSI. p ( k ) = t = 0 LN r p ( t ) r p * ( t + kD ) t = 0 LN r p (
t ) 2 , L .ltoreq. S ( 5 ) ##EQU00006##
where usually k=1.
[0038] A first decision variable d.sub.1 is computed as the sum of
the moduli of the power-normalized auto-correlations
.PSI..sub.p(k), or
d 1 = p = 1 P .PSI. p ( k ) , ##EQU00007##
where the dependency from k has intentionally been dropped.
.PSI..sub.p(k) and d.sub.1 are preferably computed continuously for
every new incoming sample.
[0039] In step 206, a transition from d.sub.1.ltoreq..phi..sub.1 to
d.sub.1>.phi..sub.1, where .phi..sub.1 is the first threshold,
generates a trigger signal to execute step 208 by the second
computational portion 110, wherein the cross-correlations of the P
received signals and the M aperiodic sequence, are computed on a
predetermined time window, sliding from -c.sub.0 to c.sub.0, as
follows:
.chi. pm ( c ) = t = .tau. - K .tau. + K r p ( t - c ) C m GD * ( t
) t = .tau. - K .tau. + K r p ( t - c ) 2 , 2 K > G D . ( 6 )
##EQU00008##
[0040] Alternatively, the following quantity can be used as
well:
.chi. . pm ( c ) = t = .tau. - K .tau. + K r p ( t - c ) C m GD * (
t ) ( t = .tau. - K .tau. + K r p ( t - c ) 2 t = .tau. - K .tau. +
K C m GD ( t ) 2 ) 1 / 2 , 2 K > G D . ( 6 ` ) ##EQU00009##
[0041] Letting c.sub.0=(1+.epsilon.)var(d.sub.1), where
0<.epsilon.<1, then the computation (6) or (6') is performed
in the interval: -c.sub.0.ltoreq.c.ltoreq.c.sub.0.
[0042] In step 210, executed by the second comparator 112, a second
decision variable d.sub.2 is computed as
d 2 ( c ) = m = 1 M p = 1 P .chi. pm ( c ) . ##EQU00010##
Packet detection is then considered achieved if
max.sub.c(d.sub.2).gtoreq..phi..sub.2, where .phi..sub.2 is the
second threshold. In the simulated implementation exemplified
below, a value for the second threshold .phi..sub.2 of 0.95 worked
well; however, .phi..sub.2 may vary in other implementations
depending, for example, on how normalization against energy is
performed.
[0043] One implementation of the invention is exemplified by way of
a cellular telecommunications environment simulated in FIGS. 3-8,
wherein the packet detection algorithm described above has been
tested in an OFDM MIMO simulator where co-channel coloured
interference and AWGN (additive white Gaussian noise) are present
to quantify the behavioural advantage with respect to the
traditional algorithm based on auto-correlation.
[0044] Accordingly, FIG. 3 presents a simulation environment 300
used for testing the packet detection algorithm, wherein
intra-system interferers are convolved with a frequency-selective
fading MIMO channel before being summed to the desired signal. The
simulation environment 300 includes an OFDM MIMO transmitter 302,
such as a base-station transmitter for next-generation cell phones,
having one or more TX antennas 304 configured for transmitting a
signal having a BW of 100 MHz, with 2048 subcarriers, via a channel
306. The channel 306 is a METRA (Multi-Element Transmit and Receive
Antenna) model based on a resampled Pedestrain-A delay profile. The
mobile speed (i.e., speed of movement of a cell phone) is 3 km/h.
The reference numeral 308 represents an AWGN generator that
introduces noise into the signal transmitted across the channel 306
at a summer 310. One or more coloured interferers 312 are
co-channel OFDM-MIMO transmitters that correspond to cell phones
used in other cells, and add interference at summers 314. The
interferers 312 use the same frame format of the transmitter 302
generating the desired signal, and the training sequences can be
all equal or user-specific. One or more RX antennas 316 are
configured for receiving the signal carried by the channel 306,
with noise and interference added thereto, and for passing the
received signal to an OFDM MIMO receiver 318. For every RX antenna
316, a total of four possible interferers are first chosen at
random, convolved with a frequency-selective fading MIMO channel,
subsequently attenuated by a random amount (min 5 dB, max 20 dB in
the simulations), and finally summed to the input signal for a
given antenna. It is understood that the RX antennas 316 and at
least a portion of the receiver 318 correspond respectively to the
two RX antennas 102 and packet detection apparatus 100 of FIG.
1.
[0045] The training sequence is generally different for every TX
antenna and follows the definition (3). FIG. 4 presents a preferred
structure of the first OFDM symbol of the training sequence output
from every TX antenna, shown together with its autocorrelation
characteristic, wherein the horizontal axis represents the samples,
and the contribution of the aperiodic portion is not included. The
first 2000 samples contain noise. Starting from sample 2001 the
periodic sequence is inserted with period 128. The first OFDM
symbol terminates at sample 4048. It can optionally be followed by
another identical symbol (without aperiodic component). Segments 8
and 9 of the first OFDM symbol have an aperiodic component summed
to the main periodic pattern. Horizontal lines show the samples
used for computing the auto-correlation (the first 1024 samples are
not used in the algorithm).
[0046] In this specific implementation S=32, D=128, .xi.=16. For
the aperiodic section: G=2, .tau.=7D. Autocorrelation is computed
with L=8. The placement of the aperiodic portion was chosen to
minimize the interference to the auto-correlation
characteristic.
[0047] FIGS. 5A, 5B, and 5C present the effect on the shape of
autocorrelation for various placements of the aperiodic part,
wherein the SNR is constant in the three cases, and the horizontal
axis represents the samples.
[0048] FIGS. 6A, 6B, 6C, 6D, and 6E exemplify how packet detection
based on auto-correlation will generate false alarms when strong
coloured co-channel interferers or a DC component are present, as
shown in FIGS. 6A, 6B, and 6D. However a narrow correlation window
after the packet detection point will reveal whether or not a
maximum is present in the cross-correlation of the aperiodic part,
so the proposed algorithm is more robust with respect to
interference. The arrow designated with a letter A indicates the
approximate time instant where the maximum of the cross-correlation
can be found. It is noted that, in practice, the cross-correlation
needs to be computed only on an interval corresponding to the
variance of threshold-crossing by auto-correlation, that is, around
300 samples or less (e.g., c.sub.0.apprxeq.150 in our
simulations).
[0049] FIG. 7 present the structure of the training sequence
indicating with different cross-hatching that sequences transmitted
from different antennas are time-orthogonal.
[0050] FIG. 8 presents performance parameters for the traditional
approach, and the proposed approach embodying features of the
present invention. On the horizontal axis, Es/No with respect to
AWGN is shown. Co-channel interference has been inserted with a
random attenuation of 5.about.20 dB, as explained above in the
simulation model.
[0051] It can be seen in FIG. 8 that for Es/No<6 dB the
performance of the auto-correlation based detection starts to drop,
indicating that in both approaches packets start to be lost. For
Es/No>6 dB on the other hand, the traditional approach starts to
be triggered on false events, and the phenomenon becomes
particularly relevant at approximately 10 dB. The proposed
approach, instead, effectively discriminates between desired
packets and interferers, showing a false-alarm rate close to zero,
which is dramatically less than the rate that may be obtained using
the traditional approach. Moreover, missed detection performance is
generally not affected by the use of a double threshold.
[0052] In cellular systems with frequency reuse 1 or higher than 1,
it is possible to have co-channel interference between packets
belonging to the user's cell and packets transmitted from other
cells. In the implementation of the invention, the discrimination
of the desired packet from the interfering packets can occur,
embedding a cell-specific code for a given cell (concept similar to
the `colour code` for GSM), that is referred to as a base-station
identifier. With the algorithm illustrated in the above paragraphs,
the use of cross-correlation will permit a cell phone to
distinguish between packets sent by different base-stations.
[0053] In the aforementioned type of implementation of the
invention, the aperiodic section of the short training symbols will
be coincident with the base-station identifier itself or one
portion of the aperiodic section will be given by the same
identifier modulated in a given way. For example, if the identifier
is a binary number, one part of the aperiodic section may contain
that identifier modulated in QPSK for PAPR reduction. The
identifier itself may be fixed for a given base station, or chosen
randomly if the network is dynamically reconfigured in time (as
could be the case for a virtual bus in the case that relay-BTS are
adopted).
[0054] By use of the present invention, the proposed approach,
including training sequence and algorithm, improves considerably
the reliability of the packet detection process. As a consequence,
useless triggering of the receiver is prevented, with a saving in
power consumption. In particular, the proposed approach has a
definite advantage in the case of co-channel coloured interference
for cellular and W-LAN systems, and in the case that a DC component
or any kind of time-periodic noise is present in the received
signal. In such cases, the proposed system with two separate
decision variables improves dramatically the false alarm
performance compared to traditional algorithms based on
auto-correlation or received energy. It should be noted here that
to optimally exploit the ability of the present invention to reject
interference, it is necessary to select the aperiodic section of
the training sequence in such a way that it is possible to
discriminate useful packets and interfering packets. The allocation
of the aperiodic section can be optimized once a given network
architecture is adopted. The "base station identifier" concept
mentioned above is one example of an optimized approach.
[0055] It is understood that the present invention may take many
forms and embodiments. Accordingly, several variations may be made
in the foregoing without departing from the spirit or the scope of
the invention. For example, if four or more TX antennas are
available, a slightly different approach can be followed. In this
approach, the periodic component of the short training symbols is
transmitted from a given subset of the total TX antennas, and the
aperiodic component of the short training symbols is transmitted
from the remaining TX antennas. Overall performance will be
comparable to the approach described above.
[0056] Having thus described the present invention by reference to
certain of its preferred embodiments, it is noted that the
embodiments disclosed are illustrative rather than limiting in
nature and that a wide range of variations, modifications, changes,
and substitutions are contemplated in the foregoing disclosure and,
in some instances, some features of the present invention may be
employed without a corresponding use of the other features. Many
such variations and modifications may be considered obvious and
desirable by those skilled in the art based upon a review of the
foregoing description of preferred embodiments. Accordingly, it is
appropriate that the appended claims be construed broadly and in a
manner consistent with the scope of the invention and that the
claims will therefore cover any such modifications or embodiments
that fall within the true scope and spirit of the invention.
* * * * *